TudCloud Inc.
All articles

NGINX Security Update: November 2026 Vulnerabilities and How to Fix Them

TudCloud Editorial7 min read
Three server nodes with a broken backend connection and an amber diagnostic warning.

Reviewed November 20, 2026. NGINX released security updates on November 18. Administrators should identify their package source, check enabled features, and deploy the patched build supplied for their installation. This guide covers NGINX Open Source; NGINX Plus, OpenResty and Kubernetes ingress products require their own vendor release guidance.

What was released?

The upstream fixed releases are NGINX 1.30.7 stable and 1.31.9 mainline. The following summary highlights issues relevant to web servers; the official list also includes mail and stream-module issues. Upstream ranges do not account for distribution backports. Source: NGINX security advisories.

CVE Issue Upstream affected range Upstream severity
CVE-2026-123007 HTTP/2 flow-control memory corruption 1.9.5–1.31.8 Major
CVE-2026-118899 Infinite loop reading a TLS 1.3 request body 1.13.0–1.31.8 Major
CVE-2026-123633 HTTP/2 use-after-free 1.19.7–1.31.8 Medium
CVE-2026-121810 HTTP/3 use-after-free 1.25.0–1.31.8 Medium
CVE-2026-123634 Resolver use-after-free 0.6.18–1.31.8 Medium
CVE-2026-125283 Buffer overflow involving alias and try_files 0.9.6–1.31.8 Minor

Fixed stable releases are exceptions to those broad numeric ranges: 1.30.7 contains the relevant fixes even though it sorts below 1.31.8. Use the advisory’s explicit fixed-version entries, not a single less-than comparison.

Does every NGINX website have the same exposure?

No. The mainline release notes describe HTTP/2 flow-control overflow as potentially crashing a worker, with other impact possible. The HTTP/2 and HTTP/3 use-after-free conditions involve insufficient worker connections. The resolver issue involves simultaneous resolution of the same name in subrequests. These are configuration and execution-path conditions, not proof that every installed server is exploitable.

The stable release notes also describe certain alias/try_files configurations, mail-proxy memory handling, and stream TLS validation issues. The mainline release additionally fixes CVE-2026-120993 affecting PROXY protocol v2 with dynamic certificate loading. Do not describe the entire batch as confirmed unauthenticated remote code execution: the official descriptions distinguish crashes, loops, validation problems and potential other impact.

Step 1: Inventory the actual installation

nginx -v
nginx -V 2>&1
sudo systemctl show nginx -p ExecStart -p MainPID
sudo nginx -t

# Debian / Ubuntu package details
dpkg-query -W 'nginx*' 2>/dev/null
apt-cache policy nginx

# RPM-based systems
rpm -qa 'nginx*'
dnf info --installed nginx

Run the package section matching your operating system. The shell’s nginx executable may not be the same one used by a panel or container. Compare ExecStart, binary paths, package release suffixes and module builds. A distribution-maintained package can retain an older upstream version while incorporating security fixes; confirm the exact installed package against that vendor’s advisory for each relevant CVE.

Inspect the effective configuration locally:

sudo nginx -T 2>&1 | less

Look for HTTP/2 settings, QUIC listeners, resolver directives, alias/try_files combinations, mail blocks and stream TLS configuration. Included files and inherited settings matter. Do not paste the complete output publicly: it can reveal internal addresses, credentials or private routing information. A text search identifies candidates for review, not a definitive vulnerability verdict.

Step 2: Reduce exposure while arranging the update

Temporary changes are not a substitute for installing fixes. The following are operational containment options based on the affected features, not a claim that one configuration workaround resolves every advisory.

  • If HTTP/2 is optional, temporarily disable it on affected listeners after checking client and gRPC requirements. In versions supporting the separate directive, use http2 off; in the appropriate context and inspect overrides. Older configurations may use the http2 parameter on listen directives. See the HTTP/2 module documentation.
  • If HTTP/3 is optional, temporarily remove the affected QUIC exposure and corresponding HTTP/3 advertisement while retaining a tested TCP HTTPS path. Verify that clients can fall back.
  • Restrict unused mail/stream endpoints to necessary peers. If a TLS request-body service cannot be patched promptly, consider routing it through a verified patched endpoint or taking the affected service out of rotation.
  • Do not globally remove resolver, alias or try_files directives without understanding dependencies. That can break upstream discovery, static files or WordPress routing.

Test configuration changes before reloading. Raising worker_connections, hiding the Server header, adding a WordPress security plugin or placing a CDN in front does not install these fixes. A CDN also leaves any reachable origin and the CDN-to-origin protocol path to be assessed separately.

Step 3: Prepare a recoverable upgrade

Keep console access available, record package/module versions and back up configuration outside the web root. For a standard /etc/nginx installation:

sudo install -d -m 700 /root/nginx-security-backups
sudo tar -czf "/root/nginx-security-backups/config-$(date +%Y%m%d-%H%M%S).tgz" -C /etc nginx
sudo nginx -t

This archive covers that directory only. Separately preserve external certificate files, panel templates, custom systemd units and container manifests as applicable. Test restoration and application compatibility in staging. A rollback to a vulnerable binary restores the security exposure; keep that endpoint isolated or mitigated while resolving the upgrade failure.

Step 4: Update through the installation’s existing owner

For ordinary OS-managed packages, refresh metadata, inspect the proposed transaction and confirm that the candidate includes the fixes before installing. Do not mix distribution and nginx.org packages casually, especially when dynamic modules are installed. Refer to NGINX’s official package guidance for upstream repositories.

Debian / Ubuntu

sudo apt update
apt-cache policy nginx
sudo apt-get -s install --only-upgrade nginx
sudo apt-get install --only-upgrade nginx

AlmaLinux / Rocky Linux / RHEL-family systems

sudo dnf --refresh info nginx
sudo dnf upgrade --assumeno 'nginx*'
sudo dnf upgrade 'nginx*'

The simulation is for review; proceed only after verifying the repository, candidate and compatible modules. Package layouts vary, so include the installed core/module packages required by your vendor. A successful command reporting “nothing to do” is not evidence of remediation when the repository has not supplied a fixed build.

For aaPanel or another panel-managed build, use its supported NGINX update mechanism and confirm the resulting binary. For containers, deploy a verified fixed image and recreate the running container; pulling an image alone changes nothing in the old instance. For custom source builds, rebuild with compatible options and modules and deploy through your established release process. Check NGINX Plus against F5’s product-specific fixed releases.

Step 5: Activate the patched binary

NGINX configuration reload and binary replacement are different operations. A reload asks the existing master to use new configuration; it does not inherently execute a newly installed binary. NGINX documents a separate live binary-upgrade procedure in Controlling nginx.

For a conventional systemd service, after a successful upgrade and configuration test, schedule a restart. This can briefly interrupt connections; clustered deployments should drain and update one instance at a time.

sudo nginx -t && sudo systemctl restart nginx
sudo systemctl is-active nginx
nginx -v
sudo systemctl show nginx -p MainPID -p ExecStart
sudo journalctl -u nginx --since '10 minutes ago' --no-pager

Verify the running master as well as the file on disk. On Linux, inspect the process selected by the service:

pid=$(systemctl show nginx -p MainPID --value)
if [ -n "$pid" ] && [ "$pid" != 0 ]; then
  sudo readlink "/proc/$pid/exe"
  sudo "/proc/$pid/exe" -v
fi

A deleted executable marker is a reason to investigate a stale process. Match the running build to the fixed upstream release or the vendor’s patched package revision; a version banner alone does not prove that every worker or container has been replaced.

Step 6: Verify the website and watch for regressions

# Replace example.com with a website you operate.
curl -fsS -o /dev/null -w '%{http_code}n' https://example.com/

Test login, uploads, reverse-proxy routes, API requests and TLS validation—not just the homepage. If a CDN fronts the site, verify the origin separately through an authorized path. Compare error rate, latency and worker restarts before and after deployment.

Worker crashes, sustained CPU use and unusual requests deserve investigation, but none alone proves exploitation. If compromise is suspected, preserve logs and system evidence before cleanup and follow an incident-response process. Installing a patch prevents the fixed vulnerability from remaining exposed; it does not undo a prior compromise.

Keep a record of the advisory date, affected features, installed package revision, activation time and validation results. For related diagnostics, read our Linux resource-pressure guide and TCP connection troubleshooting guide.